image/svg+xml
Further work
Concrete
Self sealing
Sukhbaatar Batchuluun
Supervisor: Koichiro Shiomori
Interdisciplinary Graduate School of Agriculture and Engineering, University of Miyazaki
宮崎大学
University of Miyazaki
Preparation of polystyrene microcapsules containing water droplets by solvent
evaporation method and their structural distribution analysis by machine learning
1.Introduction
2.Experimental
3.Result and discusions
4.Conclusion
...
- The structures were automatically detected by the ... and were classified by SVM.
- The structural distribution of the microcapsules prepared at a high weight ratio of solid to the organic phase and a lower ratio of that were analyzed
based on classification results.
...
Figure 3.1.1.
Microcapsule structure
Figure 3.1.2.
Comparisons of emulsion structural distribution in different conditions
Figure 3.2.1.
Diameter distributions of microcapsules.
Shelf life may
be increased
Microencapsulated
ingredients do not interfere
with other ingredients
Surface and colloidal properties
of active agents can be altered
Wider range of specific
products for consumers to
choose from
Controlled or/and
sustained release of
active agents
Liquids and gases can
be changed to pseudo solid
Protection of the active
agent from environment
Advantages
No single technique for all active
agents or product application
More skill and knowledge are
required to use this advanced
and complex technology
Possible cross-reaction that
may occur between the core
and wall material selected
Difficult to achieve
continuous and
uniform film
Disadvantages
Organic solvent
evaporation
Emulsification
Homogenisation
Filtration
Characterization
Data
Collection
Image
Segmentation
GC
BC
GR
BR
MoH
MuH
Train
Test
Feature extraction
Classification & Validation
P(r1)
<
P(r2)
P(r1)
≈
P(r2)
Mu
Mo
Production costs
To vacuum
40
0
C
To vacuum (-30kPa)
https://cvxbaatar.github.io/
https://qrgo.page.link/FZZ6o
Figure 3.2.2.
Relationships between water content and experimental condition.
Figure 3.2.3.
Relationships between diameter and experimental condition.
Figure 2.1.
Microcapsules preparation and structure classification. (Where: x
i
-feature vector, y
i
-label, p-feature vector dimension,
n-number of input points, w, b-hyperplane parameters, ξ-slack variables, C-misclassifications parameter, F-feature space.)
Crack
Microcapsule
3.1.
Structural distribution analysis by machine learning (ML)
3.2.
Preparation condition of microcapsules
ML
based measurement results
Manual
measurement results
Classifier
kNN
kNN
kNN
kNN
kNN
kNN
SVM
SVM
SVM
SVM
SVM
SVM
Accuracy, %
80.83
64.27
82.57
59.91
40.96
69.72
71.24
51.85
83.01
45.32
55.77
65.36
Table 3.1.1.
Accuracy of the proposed feature extraction techniques.
kNN- k nearest neighborhood, SVM-support vector machine
SVM was selected since it had the highest accuracy.
0.050.100.300.50
0
25
50
75
100
125
0
25
50
75
100
125
Log(Diameter)
Count
0.050.100.300.50
0
25
50
75
100
125
0
25
50
75
100
125
Log(Diameter)
Count
77% of accuracy
81% of accuracy
Further work
Concrete
Self sealing
Crack
Microcapsule
Machine
Learning
Microcapsule
1
Face
Full poster
Title
Key1
Key2
Key3
Intro1
Intro2
Intro3
Intro4
Intro5
Result1
Result2
Result3
Result4
Conclusion
Further work